So far, quite good. Nearly all fact-based queries, calculations, conversions, music selection and play instructions are understood quickly and on the first attempt. As I mentioned in my post above, though, the Echo doesn't have the same breadth of data access as Siri or Google Now, so the things you can ask for are more limited.

For example, "What's a good place for lunch around here?" returned the following:

• Siri - "I found only one restaurant around here" with a map and link to a local sandwich shop (there are many more places to eat in the same radius, though)
• Google - No verbal response, but a list of web sites that offer restaurant reviews and searches . . . no specific restaurants though
• Echo - "Hmm. I don't understand the question I heard."

The Echo transcribed the question perfectly (based on what I see logged in the companion Android app), but the back-end data store doesn't currently include that sort of location-based data. I'm sure that will change as the feature list is enhanced, though.

Working back through the request log, it appears the Echo's voice recognition is nearly perfect; where it falls down is in its access to relevant data. It will give you a news briefing (pre-recorded from NPR or BBC News), but it won't give you news on a specific topic (Siri only appears to search Twitter, though, while Google will compile a brief list of topical news items along with some general information links on the subject queried).

I think it has tremendous potential to be the information and automation nexus for the home, but it still needs significant work. That doesn't mean I find it useless, though